Cognition 150 (2016) 37–42
Contents lists available at ScienceDirect
Cognition
journal homepage: www.elsevier.com/locate/COGNIT
Short Communication
Lying despite telling the truth
Alex Wiegmann ⇑, Jana Samland, Michael R. Waldmann
Department of Psychology, University of Göttingen, Germany
a r t i c l e
i n f o
Article history:
Received 13 August 2015
Revised 22 January 2016
Accepted 26 January 2016
Keywords:
Lying
Truth
Falsity
Conversational pragmatics
Moral cognition
a b s t r a c t
According to the standard definition of lying an utterance counts as a lie if the agent believes the statement to be false. Thus, according to this view it is possible that a lie states something that happens to be
true. This subjective view on lying has recently been challenged by Turri and Turri (2015) who presented
empirical evidence suggesting that people only consider statements as lies that are objectively false
(objective view). We argue that the presented evidence is in fact consistent with the standard subjective
view if conversational pragmatics is taken into account. Three experiments are presented that directly
test and support the subjective view. An additional experiment backs up our pragmatic hypothesis by
using the uncontroversial case of making a promise.
Ó 2016 Elsevier B.V. All rights reserved.
1. Introduction
Lying is an important moral category, which has been discussed
by philosophers for centuries. Despite the great philosophical
interest in the concept of lying there are surprisingly few empirical
studies on the concept of lying coming from psychology. Psychologists have typically taken it for granted that we understand what
lying is, and have therefore addressed other issues, such as lie
detection or the developmental path of lying (e.g., Vrij, 2008; Xu,
Bao, Fu, Talwar, & Lee, 2010).
1.1. The subjective versus objective view of lying
people have an objective view on lying which means that to count
as a lie a statement has to be objectively false. The goal of this article is to defend the standard subjective view and demonstrate
empirically that the findings of Turri and Turri are consistent with
the standard view.
1.2. The studies of Turri and Turri (2015)
Turri and Turri (2015) used the following vignette in their
experiments to test whether the concept of lying entails that the
statement is objectively false1:
‘‘Making a believed-false statement to another person with the
intention that that other person believes that statement to be
true.”
[Mahon, 2008, p. 3]
‘‘Jacob’s friend Mary recently posted information on the internet that will alert the public to serious government corruption.
Soon some federal agents visit Jacob and ask where Mary is in
order to detain her. Jacob thinks that Mary is at her brother’s
house, so he tells the agents, ‘‘She is at the grocery store.” In
fact, Mary is at the grocery store.”
[p. 167]
This definition includes the so called untruthfulness condition
which states that the statement needs to be believed to be false
to constitute a lie. According to this definition it suffices that a liar
believes a statement to be false even when it is in fact true
(subjective view).
This subjective view has recently been empirically challenged
by Turri and Turri (2015). They claim to have shown that most
When participants were asked ‘‘Did Jacob lie about Mary’s location?”, the overwhelming majority (78.3%) answered with ‘‘yes,”
which supports the standard subjective view. However, Turri and
Turri (2015) claim that the responses are an artefact of the question mode. Rather than expressing whether the agent has told a
lie, they may have interpreted the test question as a request to
⇑ Corresponding author at: Department of Psychology, Gosslerstr. 14, 37073
Göttingen, Germany.
E-mail address: [email protected] (A. Wiegmann).
1
In Experiment 1 they used a slightly longer variant of this story but the quoted
version was presented in all experiments that were used to argue for the objective
view on lying.
The standard definition of lying in philosophy is:
http://dx.doi.org/10.1016/j.cognition.2016.01.017
0010-0277/Ó 2016 Elsevier B.V. All rights reserved.
38
A. Wiegmann et al. / Cognition 150 (2016) 37–42
assess whether the agent thinks he lied (i.e., perspective taking).
Alternatively, subjects may have classified the statement as a lie
because they wanted to express blame for the agent’s intent.
To control for these possible artefacts, in subsequent experiments Turri and Turri (2015) used more complex response options
that allowed subjects to separate attributions of the agent’s intent
from an assessment of whether the agent’s speech act indeed constituted a lie. In Experiment 2, the crucial options among which
subjects could choose were (out of four):
(a) ‘‘He tried to tell a lie but failed to tell a lie.”
(b) ‘‘He tried to tell a lie and succeeded in telling a lie.”
Choosing option (a) was interpreted as an endorsement of the
objective view, option (b) expresses, according to Turri and Turri,
a subjective understanding of the concept of lying. To rule out
the possible confound that subjects interpreted failing and succeeding in these response options again from the perspective of
the agent, Experiment 3 offered options (c) and (d) that were
intended to rule out this possibility. Here option (c) expresses an
objective, (d) a subjective concept of lying:
(c) ‘‘He tried to lie but only thinks he lied.”
(d) ‘‘He tried to lie and actually did lie.”
In the two experiments the overwhelming majority chose the
options (a) and (c) that are consistent with the objective view.
1.3. Conversational and experimental pragmatics
Turri and Turri (2015) interpret the findings of their Experiments 2 and 3 as supporting the objective view on lying. Here
we offer an alternative explanation of their findings that is consistent with the subjective view on lying. Our explanation is based on
conversational and experimental pragmatics (Grice, 1989; Noveck
& Reboul, 2008).
It is important to note that in both experiments of Turri and
Turri all the response options consist of two parts, which we shall
call the ‘‘trying-part” and the ‘‘result-part.” For instance, in Experiment 2 the objective option was ‘‘he tried to tell a lie but failed to
tell a lie” while the option supposed to represent the subjective
option was worded ‘‘he tried to tell a lie and succeeded in telling
a lie.” An important difference between the subjective and the
objective view is that lying is a more difficult act under the objective view compared to the subjective view. To qualify as a lie under
the subjective view, it suffices that the agent says something
believed to be false with the intent of deception. Under the objective view, these conditions hold as well, but additionally the lie
needs to state something objectively false. Because of this second
component a lie can fail under the objective view, whereas failure
is hard to conceive under the standard subjective view; here trying
to lie and lying almost never fall apart.2 In the moment the agent
has uttered something he believes to be false he has lied. No further
checks are necessary.
The ease of uttering a lie under the subjective interpretation
makes the splitting up of the response options in a trying- and a
result-part sound unnatural. When describing an action that took
its normal course and was easy to achieve, we do not split up its
description in a trying- and a result-part. An example from a different domain highlights this fact. For instance, if someone asks what
Jacob ate for lunch and he ate a hamburger, we do not reply with
2
Cases in which it may be appropriate to say that somebody tried but failed to lie
may occur in extremely unlikely scenarios. For example, somebody may try to utter a
lie but because of feelings of guilt (or some act of God) he may be unable to utter the
words.
‘‘Jacob tried to eat a hamburger and succeeded in eating a
hamburger”-we just say ‘‘Jacob ate a hamburger.” We only spilt
up the description into a trying- and a result-part when the action
was hard to perform or unlikely to be achieved (‘‘he tried to break
the world record and succeeded in breaking the world record”).
Accordingly, if the question is split up for an act that normally is
easy to accomplish, an additional unusual complication seems to
be pragmatically implied. In the eating example, the two-part format of the question may imply that there was something special
about eating the hamburger, something that goes beyond opening
the mouth and swallowing food. For example, the result-part ‘‘succeeded in eating a hamburger” could be interpreted as implying
that the whole hamburger has been eaten although it was really
too big for a normal person, or a person is described who ate the
hamburger for the first time without making a mess.
Something analogous may have happened in the lying scenario
in Turri and Turri’s experiments. When subjects who in our view
understood lying in the subjective sense were presented with the
split-up response options, they might have interpreted the question not as solely referring to whether Jacob lied. If the question
had just been about lying, less complex response options would
have been chosen (such as ‘‘he did lie” and ‘‘he did not lie”). Splitting up the answers into a trying- and a result-part might have led
subjects to conclude that the questions were not merely about the
question of whether Jacob lied but about some additional relevant
component of the story, namely the fact that what Jacob said
turned out to be true although he intended the statement to be
objectively false. Since the trying-part is identical in both response
options and only the objective option states that he failed, subjects
might have chosen the objective option not because they think that
lying requires falsity but because of the fact that this response
option expresses the failure of the agent to say something that is
objectively false.
In sum, our hypothesis is that subjects in the studies of Turri
and Turri (2015) did hold the subjective view of lying but were
led by the two-part response options to interpret the test question
not merely being about the question of whether Jacob lied but
about the question of whether what Jacob said was objectively
false. In the following experiments we will provide direct empirical
evidence for this view.
2. Experiment 1
The pragmatic implications of the split-up options used by Turri
and Turri (2015) could be avoided by providing two options, one
which states that Jacob lied and the other one stating that he did
not lie. This is exactly what Turri and Turri did in their first experiment in which the overwhelming majority stated that Jacob lied.
This finding fully supports the subjective view, and in our view is
the best test for assessing how people understand the concept of
lying. However, Turri and Turri (2015) pointed out two possible
confounds, blame and perspective taking. In the present experiment we will focus on blame. According to Turri and Turri, subjects
may have used the lying option to express their disapproval of the
protagonists’ conduct. It may certainly be true that the lie in the
scenario is considered blameworthy by people. But no empirical
evidence was offered by Turri and Turri for the further claim that
subjects actually try to express blame by changing the meaning
of the concept of lying. We remedied this deficit and actually tested
this supposition. We presented subjects either with the story used
by Turri and Turri in their Experiments 2 and 3 or with a modified
version of this story in which Jacob and Anna live in a country ruled
by a merciless dictator who commits crimes against humanity.
Anna has reported these crimes and therefore faces torture if she
gets caught. Otherwise the story was the same as in Turri and
39
A. Wiegmann et al. / Cognition 150 (2016) 37–42
Turri (2015). Jacob tells the federal agents that Anna is at a grocery
shop where in fact she is when the agents are searching for her.
Subjects were either asked whether Jacob lied (yes or no) or
whether he is to be blamed for trying to deceive the agents (yes
or no).
100%
90%
85%
80%
84%
84%
70%
67%
60%
2.1. Method
50%
40%
2.1.1. Participants
The experiment was run online in the U.K. To make sure that
subjects paid attention to the task, we asked them at the end of
the experiment whether Jacob wanted to deceive the agents and
whether he wanted to help Mary. We included the data of 249 subjects in our analyses who replied correctly to all these questions
and who worked for longer than a minute on the task. Subjects
earned 0.30£ for their participation.
30%
33%
20%
0%
original
2.2. Results and discussion
The results of the experiment are shown in Fig. 1. We replicated
the finding of the original study with the majority of subjects (85%)
declaring that Jacob lied (binomial test, 51 out of 60, test value = .5,
p < .001). Moreover, our blame manipulation was successful. Significantly more subjects chose to blame the agent in the original
story than in the dictator story (33% vs. 16%, p < .05). However,
although the blame ratings differed significantly there was no difference with regards to the question whether Jacob lied (85% vs.
84%, p = .89). These results support the subjective view (as Experiment 1 in Turri & Turri, 2015) and demonstrate that subjects did
not use the test question as an opportunity to express their disapproval of the protagonist’s conduct.
3. Experiment 2
With blame being ruled out as a confound, we now turn to our
pragmatic theory and the second confound Turri and Turri (2015)
postulated. Turri and Turri argued that in their Experiment 1 ‘‘
[. . .] participants might have simply answered in accordance with
how things seem to the agent. [. . .] Regardless of what is objectively true or false, if the agent makes a dishonest assertion, then
he thinks that he is lying. So perspective-taking could produce
the observed results.” (p. 163) To prevent subjects from taking
the protagonist’s perspective, Turri and Turri removed some details
from the story that could encourage perspective taking and provided subjects with the two-part response options, both of which
started with ‘‘he tried to lie” which reflected the perspective of
Jacob in both alternative response options. The second part then
contrasts ‘‘failed to tell a lie” with ‘‘succeeded in telling a lie”
which, according to Turri and Turri, allows subjects to express their
understanding of the concept of lying uncontaminated by the protagonist’s perspective.
Our goal of Experiment 2 was to stick as closely as possible to the
procedure proposed by Turri and Turri (2015) while blocking
the pragmatic re-interpretation of the response options. We used
the same response options as Turri and Turri in their Experiment
dictator
lie
lying/blame
2.1.2. Material and procedure
Subjects were randomly assigned to one of four conditions in a
2 (story: original vs. dictator) 2 (test question: lie vs. blame)
between-subjects design. They were presented with either the
original or the dictator story. In both conditions they were asked
‘‘Did Jacob lie about Mary’s location?’’ (yes/no) or ‘‘Is Jacob
(morally) to be blamed for trying to deceive the agents?” (yes/
no). Subsequently, subjects were asked to respond to the control
questions and asked some demographic questions.
16%
16%
15%
10%
original
dictator
blame
not lying/no blame
Fig. 1. Percentage of participants’ response choices in Experiment 1.
2 but in one condition added one additional part in which it was
stated that what was said turned out to be true. Thereby, we
intended to decrease the probability of subjects interpreting the
response options as a request to assess the protagonist’s success
or failure to state something objectively false. Given that the rest
of the response options was identical with the ones Turri and Turri
proposed, perspective taking is not an issue anymore. Moreover,
adding the obvious fact that Jacob in fact told the truth is compatible with both the subjective and the objective view. In fact, if the
objective view was true this addition should rather increase than
decrease the number of responses conforming to the objective view
because it highlights the fact that Jacob has told the objective truth.
3.1. Method
3.1.1. Participants
To make sure that the subjects paid attention to the task we
asked them at the end of the experiment whether Jacob wanted
to deceive the agents and whether he wanted to help Mary. Furthermore, to ensure that subjects paid attention to the story they
were asked whether what Jacob said about Mary’s location was
objectively true. 107 subjects who replied correctly to all these
questions and who took longer than a minute to complete the
experiment were included in the analyses. Subjects earned 0.30£
for their participation.
3.1.2. Material and procedure
Subjects were randomly assigned to either the original condition (without the new endings) or to the augmented condition
(endings in squared brackets; brackets not displayed on screen).
They were presented with the following task: ‘‘Choose the option
that better describes Jacob in the story. When Jacob spoke to the
agents about Mary’s location:
– He tried to tell a lie but failed to tell a lie [because what he said
turned out to be true].
– He tried to tell a lie and succeeded in telling a lie [although
what he said turned out to be true].”
Subsequently, subjects were asked to respond to the control
questions and were asked some demographic questions.
3.2. Results and discussion
The results of the experiment are shown in Fig. 2. We replicated
the finding of the original study with the majority of subjects (78%)
40
A. Wiegmann et al. / Cognition 150 (2016) 37–42
presented with the phrase ‘‘What Jacob said is objectively:” and
could choose between ‘‘true” and ‘‘false.” Subsequently they were
asked (in all conditions): ‘‘Which better describes Jacob?” The following response options were offered (new endings in squared
brackets; brackets not displayed on screen):
100%
90%
80%
78%
70%
72%
60%
50%
40%
30%
20%
28%
22%
10%
– He tried to lie and actually did lie [although what he said turned
out to be true].
– He tried to lie but only thinks he lied [because what he said
turned out to be true].
0%
original
original + ending
response option
lie
no lie
Fig. 2. Percentage of participants’ response choices in Experiment 2.
choosing the objective option (‘‘he tried to lie but failed in telling a
lie”) when Turri and Turri’s (2015, Experiment 2) wording was
used (binomial test, 42 out of 54, test value = .5, p < .001). In the
new augmented condition, however, in which we added the clause
that what the protagonist said turned out to be true, the majority
of subjects (72%) chose the subjective option (binomial test, 38
out of 53, test value = .5, p < .05). Hence, the two conditions
significantly differed from each other (p < .001). This finding
provides support for the view that subjects use a subjective concept of lying. The shift toward the objective option in Turri and
Turri’s framing of the response alternatives is in our view due to
a pragmatic re-interpretations induced by the framing of the
response options.
4. Experiment 3
Turri and Turri (2015) expressed doubts themselves that the
results of their Experiment 2 clearly supports the objective
view arguing that a failed lie may still be viewed as a lie. They
therefore conducted a third experiment in which they contrasted
the response options ‘‘he tried to lie but only thinks he lied”
(objective view) with the response option conforming to the
subjective view ‘‘he tried to lie and actually did lie.” We again suspected that the splitting up of the response options led subjects to
focus on the failure of the lie to state something objectively false.
To block this interpretation, we chose the same response options
as Turri and Turri (2015) but manipulated whether it was additionally mentioned that what was said turned out to be true.
Another feature of Experiment 3 of Turri and Turri (2015) is that
they added a prime immediately before the test question on the
same screen. Subjects were asked whether what Jacob said was
objectively true or false. Since we suspected that this prime may,
as a demand characteristic, have alerted subjects that in this
experiment objective truth or falsity is particularly relevant for
answering the test question, we added control conditions without
the prime in our design.
After having answered these questions, subjects were asked to
respond to the control questions and were given some demographic questions.
4.2. Results and discussion
The results are shown in Fig. 3. In the conditions in which subjects were not alerted to pay attention to the objective truth (‘‘no
prime”), we found a clear reversal. Whereas we replicated the
results supporting the objective view when Turri and Turri’s
response options were used, the pattern strongly reversed when
the new ending was added in which it was stated that what Jacob
said turned out to be true (p < .001). This reversal can also be seen
in the conditions with the prime (p < .001) although the effects
are somewhat weaker. The weakening of the effect shows that
adding the prime alerted subjects to pay more attention to the
objective truth of the utterance and strengthened the effect of
interpreting the result-part of the response option as request to
assess objective truth.
5. Experiment 4
So far we have argued that the shift toward the objective
options in Turri and Turri’s (2015) Experiments 2 and 3 are due
to a pragmatic re-interpretation of the test question induced by
the splitting of the response options in two parts. While the results
of our new experiments support this interpretation, it would be
desirable to have converging evidence for this process from a
domain in which the meaning of the target concept is not under
dispute. The act of making a promise is such a case. Promising is
viewed by linguists as a performative act which just depends on
the context of utterance and which can neither be true nor false
(Austin, 1962). Most importantly, making a promise is not invalidated by breaking the promise later. Our goal is to investigate
100%
90%
80%
81%
74%
70%
67%
60%
58%
50%
4.1. Method
40%
42%
30%
4.1.1. Participants
196 subjects met the same criteria used in the previous
experiment and earned 0.30£ for participation.
20%
33%
26%
19%
10%
0%
original
4.1.2. Material and procedure
Subjects were randomly assigned to one of the 2 (prime:
present vs. absent) 2 (ending: added vs. not added) betweensubjects conditions. In the prime conditions subjects were
original + ending
prime
lying
original
original + ending
no prime
not lying
Fig. 3. Percentage of participants’ response choices in Experiment 3.
A. Wiegmann et al. / Cognition 150 (2016) 37–42
whether using the split response options suggested by Turri and
Turri would in this case also lead to responses that apparently
seem inconsistent with the universally accepted meaning of the
concept of making a promise.
41
In the three-option condition, participants were presented with
the two response options of the two-option condition but in addition a third option was offered:
(e) ‘‘He made a promise but failed to keep it.”
5.1. Method
5.1.1. Participants
362 subjects who passed an inference attention check were
included and were paid 0.50£.
5.1.2. Material and procedure
Subjects were presented with the following story:
‘‘Jacob and Anna have been a couple for the last five years. At
the moment, however, Anna is thinking about splitting up
because Jacob has been drinking a lot of alcohol lately. In order
to save the relationship, Jacob promises Anna to never ever
touch any alcohol again. To keep his promise and thereby saving
the relationship, he throws away all the alcohol that he used to
keep in his flat. However, two weeks later, an old friend is visiting Jacob and brought, as a gift, a quite old and expensive
whiskey. Due to his promise, Jacob first denies the offer to drink
the whiskey but is not able to overcome the temptation after his
friend has asked him several times. The two get really drunk.
Because of Jacob having been drunk once again, Anna breaks
up with Jacob.”
Participants were randomly assigned to one of three conditions
that merely differed in the offered response options. In the control
condition, participants were asked to choose between two simple
options:
(a) He made a promise.
(b) He did not make a promise.
This condition resembles the options offered in Experiment 1 of
Turri and Turri (2015).
In the two-option condition, two different response options were
offered that were inspired by the two options in Turri and Turri’s
Experiment 2:
(c) ‘‘He tried to make a promise but failed to make a promise.”
(d) ‘‘He tried to make a promise and succeeded in making a
promise.”
5.1.3. Results and discussion
The results are shown in Fig. 4. The vast majority of participants
indicated that the agent in the story made a promise in both the
control condition, option (a) (binomial test, 122 out of 124, test
value = .5, p < .001) and the three-option condition, option (e)
(binomial test, 109 out of 120, test value = .33, p < .001). Most
importantly the dominant answer chosen in the experimental
two-option condition was that the agent tried but failed to make
a promise (option (c)) (binomial test, 94 out of 118, test value = .5,
p < .001).
Thus, as in the case of ‘‘lying,” participants could be pushed into
answering that the agent failed to make a promise although in the
other conditions they clearly expressed that he did make a promise. Restricting the response options apparently led to a reinterpretation of the meaning of the test question, as in the case
of lying.
6. General discussion
Our results show that Turri and Turri’s (2015) conclusion that
our concept of lying requires the statement to be objectively false
is not as clear-cut as they suggest. In Experiment 1 we showed that
their finding clearly supporting the subjective view cannot be
explained by subjects’ desire to blame the agent. In Experiments
2 and 3, the mere addition of an ending stating that the agent’s
statement turned out to be true to Turri and Turri’s original
response options also led to answers supporting the subjective
view.
We interpret these findings as showing that subjects are highly
sensitive to the pragmatic implications suggested by the response
options offered to them. When the response options seem to be at
odds with their intuitive understanding of a concept, they try to
find an interpretation that the experimenter might have intended
instead. We have shown that minimal changes can have potent
effects on subjects’ interpretation of response options. Moreover,
we demonstrated that even in the uncontroversial case of making
a promise subjects can be led to say that somebody failed to make
a promise despite the fact that with different response options they
Fig. 4. Percentage of participant’s response choices in Experiment 4.
42
A. Wiegmann et al. / Cognition 150 (2016) 37–42
could clearly declare that a promise has indeed been made. In sum,
we feel that there is no reason to abandon the standard subjective
view of lying which has been held by philosophers for centuries.
Acknowledgements
This research was supported by a grant of the Deutsche
Forschungsgemeinschaft (DFG WA 621/21-2) and Anschubfinanzierung für Nachwuchswissenschaftler/innen der Universität
Göttingen zur Vorbereitung eines Forschungsantrages. We thank
Jonas Nagel for valuable comments on an earlier draft.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.cognition.2016.
01.017.
References
Austin, J. L. (1962). How to do things with words. Oxford: Oxford University Press.
Grice, H. P. (1989). Studies in the way of words. Cambridge, MA: Harvard University
Press.
Mahon, J. E. (2008). The definition of lying and deception. In E. N. Zalta (Ed.),
Stanford encyclopedia of philosophy, Fall 2008.
Noveck, I. a., & Reboul, A. (2008). Experimental pragmatics: A Gricean turn in the
study of language. Trends in Cognitive Sciences, 12(11), 425–431. http://dx.doi.
org/10.1016/j.tics.2008.07.009.
Turri, A., & Turri, J. (2015). The truth about lying. Cognition, 138, 161–168. http://dx.
doi.org/10.1016/j.cognition.2015.01.007.
Vrij, A. (2008). Detecting lies and deceit: Pitfalls and opportunities (2nd ed.). Hoboken,
NJ: John Wiley & Sons.
Xu, F., Bao, X., Fu, G., Talwar, V., & Lee, K. (2010). Lying and truth telling in children:
From concept to action. Child Development, 81(2), 581–596.